#' Gaussian Kernel R6 class
#'
#' @docType class
#' @importFrom R6 R6Class
#' @export
#' @useDynLib GauPro, .registration = TRUE
#' @importFrom Rcpp evalCpp
#' @importFrom stats optim
# @keywords data, kriging, Gaussian process, regression
#' @return Object of \code{\link[R6]{R6Class}} with methods for fitting GP model.
#' @format \code{\link[R6]{R6Class}} object.
#' @examples
#' k1 <- Gaussian$new(beta=0)
#' plot(k1)
#' k1 <- Gaussian$new(beta=c(0,-1, 1))
#' plot(k1)
#'
#'
#' n <- 12
#' x <- matrix(seq(0,1,length.out = n), ncol=1)
#' y <- sin(2*pi*x) + rnorm(n,0,1e-1)
#' gp <- GauPro_kernel_model$new(X=x, Z=y, kernel=Gaussian$new(1),
#' parallel=FALSE)
#' gp$predict(.454)
#' gp$plot1D()
#' gp$cool1Dplot()
Gaussian <- R6::R6Class(classname = "GauPro_kernel_Gaussian",
inherit = GauPro_kernel_beta,
public = list(
# initialize = function(beta, s2=1, beta_lower=-8, beta_upper=6,
# s2_lower=1e-8, s2_upper=1e8) {
# self$beta <- beta
# self$beta_length <- length(beta)
# # if (length(theta) == 1) {
# # self$theta <- rep(theta, self$d)
# # }
# self$beta_lower <- beta_lower
# self$beta_upper <- beta_upper
#
# self$s2 <- s2
# self$logs2 <- log(s2, 10)
# self$logs2_lower <- log(s2_lower, 10)
# self$logs2_upper <- log(s2_upper, 10)
# },
#' @description Calculate covariance between two points
#' @param x vector.
#' @param y vector, optional. If excluded, find correlation
#' of x with itself.
#' @param beta Correlation parameters.
#' @param s2 Variance parameter.
#' @param params parameters to use instead of beta and s2.
k = function(x, y=NULL, beta=self$beta, s2=self$s2, params=NULL) {
if (!is.null(params)) {
lenparams <- length(params)
if (self$beta_est) {
beta <- params[1:self$beta_length]
} else {
beta <- self$beta
}
if (self$s2_est) {
logs2 <- params[lenparams]
} else {
logs2 <- self$logs2
}
s2 <- 10 ^ logs2
} else {
if (is.null(beta)) {beta <- self$beta}
if (is.null(s2)) {s2 <- self$s2}
}
theta <- 10^beta
if (is.null(y)) {
if (is.matrix(x)) {
# cgmtry <- try(val <- s2 * corr_gauss_matrix_symC(x, theta))
# if (inherits(cgmtry,"try-error")) {browser()}
# return(val) # arma version isn't actually faster?
return(s2 * corr_gauss_matrix_symC(x, theta))
# return(s2 * corr_gauss_matrix_sym_armaC(x, theta))
} else {
return(s2 * 1)
}
}
if (is.matrix(x) & is.matrix(y)) {
s2 * corr_gauss_matrixC(x, y, theta)
# if (self$D >= 12 || nrow(x) < 30) {
# s2 * corr_gauss_matrixC(x, y, theta)
# } else { # parallel only faster for small D and many rows
# s2 * corr_gauss_matrixCpar(x, y, theta)
# }
# s2 * corr_gauss_matrix_armaC(x, y, theta) # arma not actually faster?
# corr_gauss_matrix_armaC(x, y, theta, s2)
} else if (is.matrix(x) & !is.matrix(y)) {
s2 * corr_gauss_matrixvecC(x, y, theta)
} else if (is.matrix(y)) {
s2 * corr_gauss_matrixvecC(y, x, theta)
} else {
s2 * exp(-sum(theta * (x-y)^2))
}
},
#' @description Find covariance of two points
#' @param x vector
#' @param y vector
#' @param beta correlation parameters on log scale
#' @param theta correlation parameters on regular scale
#' @param s2 Variance parameter
kone = function(x, y, beta, theta, s2) {
if (missing(theta)) {theta <- 10^beta}
s2 * exp(-sum(theta * (x-y)^2))
},
#' @description Derivative of covariance with respect to parameters
#' @param params Kernel parameters
#' @param X matrix of points in rows
#' @param C_nonug Covariance without nugget added to diagonal
#' @param C Covariance with nugget
#' @param nug Value of nugget
dC_dparams = function(params=NULL, X, C_nonug, C, nug) {
n <- nrow(X)
lenparams <- length(params)
if (lenparams > 0) {
if (self$beta_est) {
beta <- params[1:self$beta_length]
} else {
beta <- self$beta
}
if (self$s2_est) {
logs2 <- params[lenparams]
} else {
logs2 <- self$logs2
}
} else {
beta <- self$beta
logs2 <- self$logs2
}
# lenparams <- length(params)
# beta <- params[1:(lenparams - 1)]
theta <- 10^beta
log10 <- log(10)
# logs2 <- params[lenparams]
s2 <- 10 ^ logs2
# if (inherits(try(diag(nug*s2, nrow(C_nonug))), "try-error")){browser()}
# if (is.null(params)) {params <- c(self$beta, self$logs2)}
if (missing(C_nonug)) { # Assume C missing too, must have nug
C_nonug <- self$k(x=X, params=params)
C <- C_nonug + diag(nug*s2, nrow(C_nonug))
}
lenparams_D <- self$beta_length*self$beta_est + self$s2_est
# I wrote Rcpparmadillo function to speed this up a lot hopefully
# useR <- FALSE
if (!self$useC) { # useR
dC_dparams <- array(dim=c(lenparams_D, n, n), data=0)
if (self$s2_est) {
dC_dparams[lenparams_D,,] <- C * log10 #/ s2 * s2 *
}
# dC_dparams <- rep(list(C_nonug), length(beta))
if (self$beta_est) {
for (k in 1:length(beta)) {
for (i in seq(1, n-1, 1)) {
for (j in seq(i+1, n, 1)) {
# if (inherits(try(C_nonug[i,j] * (X[i,k] - X[j,k])^2 *
# theta[k] * log10), "try-error")) {browser()}
dC_dparams[k,i,j] <- - C_nonug[i,j] * (X[i,k] - X[j,k])^2 *
theta[k] * log10
dC_dparams[k,j,i] <- dC_dparams[k,i,j]
}
}
for (i in seq(1, n, 1)) { # Get diagonal set to zero
dC_dparams[k,i,i] <- 0
}
}
}
} else {
dC_dparams <- kernel_gauss_dC(X, theta, C_nonug, self$s2_est,
self$beta_est, lenparams_D, s2*nug)
}
# mats <- c(dC_dbetas, list(dC_dlogs2))
return(dC_dparams)
},
#' @description Calculate covariance matrix and its derivative
#' with respect to parameters
#' @param params Kernel parameters
#' @param X matrix of points in rows
#' @param nug Value of nugget
C_dC_dparams = function(params=NULL, X, nug) {
n <- nrow(X)
lenparams <- length(params)
if (lenparams > 0) {
if (self$beta_est) {
beta <- params[1:self$beta_length]
} else {
beta <- self$beta
}
if (self$s2_est) {
logs2 <- params[lenparams]
} else {
logs2 <- self$logs2
}
} else {
beta <- self$beta
logs2 <- self$logs2
}
# if (is.null(params)) {params <- c(self$beta, self$logs2)}
# beta <- params[1:(lenparams - 1)]
theta <- 10^beta
log10 <- log(10)
# logs2 <- params[lenparams]
s2 <- 10 ^ logs2
# Calculate C
C_nonug <- self$k(x=X, beta=beta, s2=s2)
C <- C_nonug + diag(nug*s2, nrow(C_nonug))
lenparams_D <- self$beta_length*self$beta_est + self$s2_est
# I wrote Rcpparmadillo function to speed this up a lot hopefully
# useR <- FALSE
if (!self$useC) { # useR
dC_dparams <- array(dim=c(lenparams_D, n, n), data=0)
if (self$s2_est) {
dC_dparams[lenparams_D,,] <- C * log10 #/ s2 * s2 *
}
# dC_dbetas <- rep(list(C_nonug), length(beta))
# n <- nrow(X)
if (self$beta_est) {
for (k in 1:length(beta)) {
for (i in seq(1, n-1, 1)) {
for (j in seq(i+1, n, 1)) {
dC_dparams[k,i,j] <- - C[i,j] * (X[i,k] - X[j,k])^2 *
theta[k] * log10
dC_dparams[k,j,i] <- dC_dparams[k,i,j]
}
}
for (i in seq(1, n, 1)) { # Get diagonal set to zero
dC_dparams[k,i,i] <- 0
}
}
}
} else {
dC_dparams <- kernel_gauss_dC(X, theta, C_nonug, self$s2_est,
self$beta_est, lenparams_D, s2*nug)
}
# kernel_gauss_dC(X, theta, C_nonug, self$s2_est,
# self$beta_est, lenparams_D, s2*nug)
# mats <- c(dC_dbetas, list(dC_dlogs2))
return(list(C = C, dC_dparams = dC_dparams))
},
# dC_dx = function(XX, X, theta, beta=self$beta, s2=self$s2) {
# if (missing(theta)) {theta <- 10^beta}
# if (!is.matrix(XX)) {stop("XX must be matrix")}
# d <- ncol(XX)
# if (ncol(X) != d) {stop("XX and X must have same number")}
# n <- nrow(X)
# nn <- nrow(XX)
# dC_dx <- array(NA, dim=c(nn, d, n))
# for (i in 1:nn) {
# for (j in 1:d) {
# for (k in 1:n) {
# dC_dx[i, j, k] <- -2 * theta[j] * (XX[i, j] - X[k, j]) *
# s2 * exp(-sum(theta * (XX[i,] - X[k,]) ^ 2))
# }
# }
# }
# dC_dx
# },
# Below is updated version using arma, was called dC_dx_arma before
#' @description Derivative of covariance with respect to X
#' @param XX matrix of points
#' @param X matrix of points to take derivative with respect to
#' @param theta Correlation parameters
#' @param beta log of theta
#' @param s2 Variance parameter
dC_dx = function(XX, X, theta, beta=self$beta, s2=self$s2) {
if (missing(theta)) {theta <- 10^beta}
if (!is.matrix(XX)) {stop("XX must be matrix")}
if (ncol(X) != ncol(XX)) {stop("XX and X must have same number of cols")}
corr_gauss_dCdX(XX, X, theta, s2)
},
#' @description Second derivative of covariance with respect to X
#' @param XX matrix of points
#' @param X matrix of points to take derivative with respect to
#' @param theta Correlation parameters
#' @param beta log of theta
#' @param s2 Variance parameter
d2C_dx2 = function(XX, X, theta, beta=self$beta, s2=self$s2) {
if (missing(theta)) {theta <- 10^beta}
if (!is.matrix(XX)) {stop("XX must be matrix")}
d <- ncol(XX)
if (ncol(X) != d) {stop("X and XX must have same # of columns")}
n <- nrow(X)
nn <- nrow(XX)
d2C_dx2 <- array(NA, dim=c(nn, d, d, n))
for (i in 1:nn) {
for (k in 1:n) {
Cik <- s2 * exp(-sum(theta * (XX[i,] - X[k,]) ^ 2))
if (d > 1) {
for (j1 in 1:(d-1)) {
for (j2 in (j1+1):d) {
d2C_dx2[i, j1, j2, k] <- 4 * theta[j1] *
(XX[i, j1] - X[k, j1]) * theta[j2] *
(XX[i, j2] - X[k, j2]) * Cik
d2C_dx2[i, j2, j1, k] <- d2C_dx2[i, j1, j2, k]
}
}
}
for (j in 1:d) {
d2C_dx2[i, j, j, k] <- -2 * theta[j] * Cik +
4 * theta[j]^2 * (XX[i, j] - X[k, j])^2 * Cik
}
}
}
d2C_dx2
},
#' @description Second derivative of covariance with respect to
#' X and XX each once.
#' @param XX matrix of points
#' @param X matrix of points to take derivative with respect to
#' @param theta Correlation parameters
#' @param beta log of theta
#' @param s2 Variance parameter
d2C_dudv = function(XX, X, theta, beta=self$beta, s2=self$s2) {
if (missing(theta)) {theta <- 10^beta}
if (!is.matrix(XX)) {stop("XX must be matrix")}
d <- ncol(XX)
if (ncol(X) != d) {stop("X and XX must have same # of columns")}
n <- nrow(X)
nn <- nrow(XX)
d2C_dx2 <- array(NA, dim=c(nn, d, d, n))
for (i in 1:nn) {
for (k in 1:n) {
Cik <- s2 * exp(-sum(theta * (XX[i,] - X[k,]) ^ 2))
if (d > 1) {
for (j1 in 1:(d-1)) {
for (j2 in (j1+1):d) {
d2C_dx2[i, j1, j2, k] <- - 4 * theta[j1] *
(XX[i, j1] - X[k, j1]) * theta[j2] *
(XX[i, j2] - X[k, j2]) * Cik
d2C_dx2[i, j2, j1, k] <- d2C_dx2[i, j1, j2, k]
}
}
}
for (j in 1:d) {
d2C_dx2[i, j, j, k] <- 2 * theta[j] * Cik -
4 * theta[j]^2 * (XX[i, j] - X[k, j])^2 * Cik
}
}
}
d2C_dx2
},
#' @description Second derivative of covariance with respect to X and XX
#' when they equal the same value
#' @param XX matrix of points
#' @param theta Correlation parameters
#' @param beta log of theta
#' @param s2 Variance parameter
d2C_dudv_ueqvrows = function(XX, theta, beta=self$beta, s2=self$s2) {
# Calculates derivative of C w.r.t. each component evaluated for
# both components equal to rows of XX
# Vectorized version of d2C_dudv for u=v for rows of XX
# Name is for "u equal v for rows of XX"
# Much simpler since XX-X terms go to zero when XX=X
# For m1 matrix, following two are equal, this version 2.5x faster
# lapply(1:nrow(m1), function(i) {gp$kernel$d2C_dudv(XX = m1[i,,drop=F],
# X = m1[i,,drop=F])[1,,,1]})
# gp$kernel$d2C_dudv_ueqvrows(XX = m1)
if (missing(theta)) {theta <- 10^beta}
if (!is.matrix(XX)) {stop("XX must be matrix")}
d <- ncol(XX)
nn <- nrow(XX)
d2C_dx2 <- array(0, dim=c(nn, d, d))
for (j in 1:d) {
# Not multiplied by C since C=1 when u=v
d2C_dx2[, j, j] <- 2 * theta[j] * s2
}
d2C_dx2
},
#' @description Print this object
print = function() {
cat('GauPro kernel: Gaussian\n')
cat('\tD =', self$D, '\n')
cat('\tbeta =', signif(self$beta, 3), '\n')
cat('\ts2 =', self$s2, '\n')
}
)
)
#' @rdname Gaussian
#' @export
#' @param beta Initial beta value
#' @param s2 Initial variance
#' @param D Number of input dimensions of data
#' @param beta_lower Lower bound for beta
#' @param beta_upper Upper bound for beta
#' @param beta_est Should beta be estimated?
#' @param s2_lower Lower bound for s2
#' @param s2_upper Upper bound for s2
#' @param s2_est Should s2 be estimated?
#' @param useC Should C code used? Much faster.
k_Gaussian <- function(beta, s2=1, D,
beta_lower=-8, beta_upper=6, beta_est=TRUE,
s2_lower=1e-8, s2_upper=1e8, s2_est=TRUE,
useC=TRUE) {
Gaussian$new(
beta=beta,
s2=s2,
D=D,
beta_lower=beta_lower,
beta_upper=beta_upper,
beta_est=beta_est,
s2_lower=s2_lower,
s2_upper=s2_upper,
s2_est=s2_est,
useC=useC
)
}
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.